Introduction

Valorant is a free-to-play first-person hero shooter developed and published by Riot Games, for Microsoft Windows. First teased under the codename Project A in October 2019, the game began a closed beta period with limited access on April 7, 2020, followed by an official release on June 2, 2020.

The weapons dataset is based on their first ever major tournament, Stage 2: "Masters" of the VCT (Valorant Champions Tour) 2021 which took place between 24th May and 30th May in Iceland.

And the new dataset is taken from the game directly (patch 4.04)

from IPython.display import HTML HTML('''

Types of weapons

Models :

What we can see here us that the damages (Head, body and leg ) are correlated between each other. The price is kind of correlated with the damage and the magazine capacity (with some outliers). And the fire rate is kind of independent from the rest.

First we try with categorical variables, then without categorical variables.

The dataset set is small so there isn't much to draw from it but we are just playing

Let's see some cases

The price is positively correlated witht the head damage, but it is weird that's the coefficient is negative for the body damage. The spead is negative as expected, (HIP less infulential), and the fire rate has a negative coeff too. The category is having a considerable effect too.

Without scaling, the categorical coeff has more value, which expected (as an equilibrium with the previous case)

Now without the categories

It's weird that the Leg damage has a relatively big coefficient (in the absolute value), and that the HIP has a positive coefficient. Note that while the error is big the performance of the model is better than without the spread variables.

All the errors are huge, this is not something to model with linear regression , also other variables aren't taken into consideration , like the spray.

Matches :

The dataset here is taken from here on kaggle and it's taken from vlr.gg .

Content :

There is four tables. The top level is Matches that will tell you teams playing and match (map) score. Game is the next level that breaks down the specific maps played. Then GameRounds gives a round by round breakdown which shows who won, economy of each team, win type, and buy type, whenever the info is available. The game rounds are packaged in one string that you should be able to cast as a json. Lastly there is GameScoreboard which gives you the player performance, as well as things like number of first kills, first deaths, 2Ks, 3Ks, One v Ones, One v Twos, ect.

This content introduction is made by Joshua Broas on kaagle.

Agents :